aeolusyansheng/qdrant-knowledge-base

fix: add langchain-google-vertexai shim for ragas 0.2.15 compatibility

feat: load GCP credentials from GOOGLE_CREDENTIALS_JSON secret on HF Space

fix: use google-auth ADC for Vertex AI OAuth2 token instead of API key

feat: migrate LLM from Groq to Gemini 2.5 via Vertex AI OpenAI-compatible endpoint

fix: localize doc_type tag in source cards

feat: support cross-language doc_type filtering using MatchAny

fix: correctly handle 'All' doc_type filter across multiple languages

fix: localize hardcoded error messages

feat: sync response language with UI language

Improve Japanese documentation and comments, change default language to English

Add Japanese version of README (README_JP.md) with half-width katakana

Fix Japanese text to use half-width katakana and add doc type translation in sidebar

Fix file uploader: make file selection button functional while keeping custom styling

Add dynamic index status check with retry functionality

Improve document list readability with card-style layout and spacing

Sidebar UI improvements: update status indicator, add author info, adjust spacing

UI improvements: remove file icon and style doc_type as tag like framework

Use custom side-label class for framework and doc_type labels to match upload section styling

Add more specific CSS selectors for sidebar labels to ensure style override

Unify sidebar label font styles to match upload section header

Update sync_to_hf.yml workflow

docs: update README

ci: add GitHub Actions sync to HuggingFace

fix: inline styles.css tokens into app.py to remove external file dependency

feat: redesign UI with custom chat bubbles, language switcher, and polished layout

feat: add RAGAS evaluation with Groq backend

feat: initial commit with hybrid search, reranker, markdown support and document management